Glyceraldehyde Phosphate Dehydrogenase of Escherichia coli STRUCTURAL AND CATALYTlC PROPERTIES*

SUMMARY The molecular weight of the Escherichia coli glyceraldehyde phosphate dehydrogenase subunit in 5 M guanidine hydrochloride containing 0.01 M dithiothreitol was determined by high speed sedimentation equilibrium to be 35,000. Amino acid composition of the protein, taken in conjunction with the number of peptides obtained after tryptic digestion, indicated that the subunits of the enzyme were very probably identical, each containing 12 arginine and 26 lysine residues. NAD+ content of the protein, measured both by spectrophotometric and enzymatic methods, was found to be very low (~0.1 mole of NADf per 35,000 g of protein), although capacity to strongly bind NAD+ was intact. The enzyme could be crystallized but only after addition of NAD+ to apoprotein. when


SUMMARY
The molecular weight of the Escherichia coli glyceraldehyde phosphate dehydrogenase subunit in 5 M guanidine hydrochloride containing 0.01 M dithiothreitol was determined by high speed sedimentation equilibrium to be 35,000. Amino acid composition of the protein, taken in conjunction with the number of peptides obtained after tryptic digestion, indicated that the subunits of the enzyme were very probably identical, each containing 12 arginine and 26 lysine residues. NAD+ content of the protein, measured both by spectrophotometric and enzymatic methods, was found to be very low (~0.1 mole of NADf per 35,000 g of protein), although capacity to strongly bind NAD+ was intact.
The enzyme could be crystallized but only after addition of NAD+ to apoprotein. One cysteine residue per subunit was carboxymethylated with resultant enzymatic inactivation when native enzyme was reacted with iodoacetic acid; similarly there was one sulfhydryl group per subunit which was rapidly reactive with 5,5'-dithiobis(2-nitrobenzoic acid). Under protein denaturing conditions a total of 3 cysteine residues per subunit were detected although analyses of performic acid-oxidized protein indicated four cysteic acids per subunit. Dehydrogenase activity of the enzyme was markedly sulfhydryl dependent; there was <25% of maximal activity if sulfhydryl compounds were not included in assay mixtures. Esterase activity (with p-nitrophenyl acetate as substrate) was inhibited by AMP, ADP, and ATP but not by NAD+.
In the preceding paper a scheme was described for purification of three Embden-Meyerhof pathway enzymes from extracts of Escherida coli (1). One of these, glyceraldehyde phosphate dehydrogenase (n-glyceraldehyde 3-phosphate: NAD+ oxidoreductase, phosphorylating, EC 1.2.1.12), was obtained in homogeneous form and found to have physical properties similar to those of analogous proteins from other sources in nature.
Although previous work by Allison  detail because of widespread interest in the structural, catalytic, and allosteric properties of glyceraldehyde phosphate dehydrogenases in general (4, 5) and because of the importance of all E. coli proteins as components of a model cell whose biochemistry has been extensively studied. A preliminary report of these studies has appeared (6).
Operations were conducted at 2", unless otherwise indicated, and preparative centrifugations were performed at 15,000 x g for 15 min.
Triethylamine was redistilled before use, and triethylammonium bicarbonate buffer, pH 7.5, was prepared as described previously (7). The p-nitrophenyl acetate and pnitrophenol were recrystallized twice from chloroform-heptane mixtures (melting points, 78.5" and 115.5", respectively Performic acid-oxidized protein was prepared by exhaustive dialysis uersus 10 mM triethylammonium bicarbonate buffer, pH 7.5, followed by lyophilization and treatment with performic acid for 4 hours at 0" according to Hirs (8). Excess performic acid was removed by lyophilization (9). S-Carboxymethylation of protein was carried out at pH 8, with a 3-to 8-fold molar excess of iodoacetic acid over total protein sulfhydryl.
Either 5 M guanidine hydrochloride or 8 M urea was used to prepare denatured carboxymethylated protein, and 2-mercaptoethanol (amount equimolar with iodoacetic acid) was included when reducing conditions were desired (10). Details are given in Table III.
Charcoal-treated enzyme was prepared by addition of 50 mg of dried Norit A to a solution containing 6 mg of protein in 0.5 ml of Tris-EDTA buffer.
After the suspension had been mixed and incubated at 2" for 2 hours, it was centrifuged, and the precipitate was resuspended in 1 ml of Tris-EDTA buffer. After recentrifugation both supernatants were combined and clarified by a final centrifugntion. NAD+-loaded enzyme was obtained by addition of NAD+ (10 moles per 144,000 g of protein) to a solution containing 8 to 12 mg of enzyme per ml of 0.8 M neutral ammonium sulfate in Tris-EDTA buffer.
After the mixture had incubated for 20 min at 2", 16 volumes of saturated neutral ammonium sulfate were added, and incubation was continued for an additional 16 hours.
The precipitate was collected by centrifugation, suspended in 1 volume of saturated neutral ammonium sulfate, centrifuged again, and finally dissolved in 1 volume of Tris-EDTA buffer.
(Neutral saturated ammonium sulfate was prepared as described in the preceding paper (I).) Sedimenlation Studies-Molecular weight of the protein subunits was determined with the high speed sedimentation equilibrium technique of Yphantis (11)) as described in the preceding paper (1). Reduced, denatured protein was prepared by dialysis at 23" against a solvent containing 5 M guanidine hydrochloride, 0.01 M dithiothreitol, 0.1 M NaCl, 0.01 M sodium phosphate, pH 7, according to a protocol outlined previously (12), and centrifuge cells were filled with Hamilton syringes immediately after the dialysis sacs had been opened. Sedimentation analyses were carried out at 23.8" for 39 hours with a rotor speed of 40,360 rpm.
Partial specific volume of the denatured glyceraldehyde phosphate dehydrogenase polypeptide chains was assumed to be 4' = 0.723 cc per g, a value 98.5% that of the native E. coli protein (Table I) (13,14). Solvent density was determined by pycnometry to be 1.123 g per cc. When commercial rabbit muscle glyceraldehyde phosphate dehydrogenase was examined with the same procedure, subunit molecular weight results were identical with those described below for the E. coli enzyme.
Other Methods-Amino acid analyses were generously provided by Dr. George R. Stark, Stanford University. Spectrophotometric estimates of tryptophan content were made according to the method of Goodwin and Morton (15), Bencze and Schmid (16), and Edelhoch (17)  based upon these results the extinction coefficient of purified enzyme in Tris-EDTA buffer was 1000 cm*g-1 at 280 rnp. Glyceraldehyde phosphate dehydrogenase activity was assayed as described in the preceding paper (1). Esterase activity was determined spectrophotometrically at 23" in a mixture containing 20 mM N-tris(hydroxymethyl)methyl glycine buffer, pH 7.8, and 1 mM p-nitrophenyl acetate (26); corrections were made for spontaneous hydrolysis of the substrate.
The extinction coefficient of liberated p-nitrophenol under these conditions was verified with authentic, recrystallized compound (E = 16,500 Rr-lcm-l at 400 mp).
Absorption spectra were recorded on a Cary model 15 spectrophotometer, and radioactivity measurements were made in a Packard scintillation detector.

RESULTS
Structural Properties Subunit Structure of Enzyme-In the preceding paper the sedimentation properties of native E. coli glyceraldehyde phosphate dehydrogenase were examined; the molecular weight, determined during high speed sedimentation equilibrium, was 144,000 (mean) f 3,900 (S.D.), and the sedimentation coefficient was s:~,~ = 7.5 S (1). These figures are very similar to corresponding values obtained from studies of glyceraldehyde phosphate dehydrogenase proteins isolated from a variety of sources (2, 4). When the subunit structure of such proteins has been examined, the results have invariably shown four polypeptide chains per native enzyme molecule (4, 27, 28)) and in the cases of the pig and lobster muscle enzymes, amino acid sequence determinations have shown the subunits to be identical (4). The molecular weight of reduced denatured E. coli glyceraldehyde phosphate dehydrogenase was determined during high speed sedimentation equilibrium in a solvent environment of 5 M guanidine hydrochloride and 0.01 M dithiothreitol. The results of three analyses with initial concentrations ranging from 0.5 to 2.6 mg of protein per ml of solvent were: Mwspp (mean) = 35,200 f 100 (S.D.); Range: 35,400 where MwaPP represents the apparent weight-average molecular weight of the macromolecule obtained from linear least squares slopes of log (fringe number) versus (radius)2 plots of the data. As in studies of the native enzyme (l), such plots were highly linear, a finding suggestive of subunit size homogeneity.
b See Table III. c The value for tryptophan was obtained from separate spectrophotometric and chemical analyses; see the text. d The mean residue weight calculated from these data is 107.9; the calculated partial specific volume of the protein is t = 0.734 cc per g (33).
it appears that the E. coli enzyme conforms to the general pat- densely staining, reproducible peptides are outlined with a solid line, and faintly staining or inconstant peptides are outlined with a dashed line. Selective staining procedures for arginine-containing peptides (marked A) and histidine-containing peptides (marked H) were carried out as described in Reference 7.
acid-treated preparations of this enzyme is given in Table I along with the earlier data of Allison and Kaplan; it can be seen that, except for tryptophan and, to a lesser extent, arginine, serine, alanine, and methionine, there is good agreement between the two sets of data. The cysteic acid content (15.5 moles per mole of enzyme) was obtained from performic acid-oxidized protein. Reactivity of cysteines with iodoacetic acid is described in a later section. Analyses for tryptophan by three spectrophotometric methods yielded estimates of 15.8 (15), 15.9 (16), and 16.5 (17) moles per mole of enzyme, in agreement with results obtained by means of a chemical method (16.2 moles of tryptophan per mole of enzyme) (18).
If the four polypeptide subunits of this enzyme are identical, their composition is indicated in the fourth column of Table I.
Peptid,e Maps of Tryptic Digests- Fig. 1 shows a typical peptide map of a tryptic digest of performic acid-oxidized E. coli glyceraldehyde phosphate dehydrogenase. The expected number of tryptic fragments is 39 if the subunits are identical and each contains 26 lysine and 12 arginine residues (Table I). In all maps the number of ninhydrin-positive spots was in the range of 36 to 40. Of these, 10 peptides stained positively for arginine 10 Curve a, isolated native enzyme; Curve b, charcoal-treated enzyme; Curve c, NAD+-loaded enzyme.
The ordinate represents the extinction of a 1 g per 100 ml of solution of protein over an optical path of 1 cm, and the abscissa is the wave length of measurement. and 5 stained positively for histidine.
From Table I the expectation is 12 arginine-and 6 histidine-containing tryptic fragments. Over-all, these results are in good accord with prediction if the four subunits of the enzyme are identical. NAD+ Content-The ultraviolet absorption spectrum of E. coli glyceraldehyde phosphate dehydrogenase is shown in Curve a of Fig. 2. The X,,, (278 nm) and the ratio 280: 260 = 1.6 suggested a very low NAD+ content (34).
However, a small amount of coenzyme was probably present because when the enzyme was treated with charcoal, the ratio 280:260 rose to 1.8, although the x max remained at 278 mp (Curve b of Fig. 2). The enzyme was then NAD+-loaded as described under "Experimental Procedure." After this treatment the X,, was shifted to 276 rnp, the ratio 280:260 fell to 1.15, and a broad absorption band appeared between 300 and 400 rnp (Curve c of Fig. 2). This band is a characteristic feature of NAD+-glyceraldehyde phosphate dehydrogenase complexes isolated from several mammalian and crustacean species (2, 35). Titration of the E. coli enzyme with NAD+ could be followed at 360 rnp (Fig. 3). There was a linear increase in absorbance up to the addition of about 3.5 moles of NADf per mole of protein; then the slope changed, and only a very small change in absorbance was observed upon further addition of coenzyme.
The NAD+ contents of the enzyme, both as isolated and after NADf-loading, were directly determined according to the method of Villee (19). Parallel analyses of rabbit muscle enzyme were simultaneously carried out as a control.
Results are summarized in Table II, where it can be seen that the E. coli enzyme contained less than 0.3 mole of NAD+ per mole of native protein; by contrast, rabbit muscle enzyme was found to have 1.9 moles of NAD+ per mole of protein.
After NAD+ loading both enzymes contained nearly 4 moles of coenzyme per mole of protein.
Once the E. co& enzyme was NAD+-loaded, the pyridine nucleotide appeared to bind very tightly. Passage of NAD+loaded enzyme over DEAF-Sephadex (A-50) or Sephadex G-200 columns (as in the preparation of Fractions V-DH and VI-DH, respectively, in the purification procedure cited in the preceding paper (1) which continued to resemble Curve c of Fig. 2 with a ratio 280: 260 = 1.2. A similar result was obtained after 48 hours of dialysis of NAD+-loaded enzyme versus 2000 volumes of Tris-EDTA buffer at 0". We have not undertaken an analysis of the association constants of the four bound NAD+ residues, and it is likely that small amounts of the most loosely bound NAD+ ($36) were removed by these treatments.
Nevertheless, the bulk of the coenzyme remained closely associated with the protein.  (Fig. 4). Suljhydryl Reactivity-The 4 half-cystine residues detected in the performic acid-oxidized enzyme subunit (Table I) had markedly different sulfhydryl reactivities in analyses with iodoacetic acid and 5,5'-dithiobis(2-nitrobenzoic acid). As shown in Table  III (No. l), native or NAD+-loaded enzyme was inactivated by interaction with 5.3 moles of iodoacetic acid (approximately one carboxymethylated cysteine per enzyme subunit). Similarly there was approximately 1 cysteine residue per subunit which reacted rapidly with 5,5'-dithiobis(2-nitrobenzoic acid). This residue presumably corresponds to cysteine-149 in the active center of pig and lobster muscle glyceraldehyde phosphate dehydrogenases, the residue which is selectively carboxymethylated by iodoacetic acid and which is selectively acylated during enzymatic esterolysis (4). The comparative studies of Allison have shown that the E. coli enzyme has the same active site peptide sequence as the other proteins, including this particular reactive cysteine (3). During the slow phase of 5,5'-dithiobis(2nitrobenzoic acid) interaction with native enzyme a second subunit cysteine became reactive (No. 1 of Table III). This is presumed to correspond to cysteine-153 in the active site peptide, just 4 residues removed from the rapidly reacting cysteine.
The mechanism of release of the second colored thionitrobenzoate ion may be due to formation of a disulfide bond between cysteines 149 and 153, according to the proposal of Wasserman and Major (37). In the denaturing environment of 5 M guanidine hydrochloride or 8 Y urea two cysteines per subunit were reactive with iodoacetic acid, one of these corresponding to cysteine-149 since the protein was previously incubated with iodoacetic acid prior to addition of the denaturing agent; however, 5,5'-dithiobis(2nitrobenzoic acid) analysis revealed 3 reactive residues (No. 2 of a Enzyme (0.03 pmole) was incubated at 23' with iodoacetic-l-14C acid (4 coles, 5 X 10' dpm) in 2 ml of Buffer A. After 5 min aliquots were assayed for catalytic activity; <5'% of original activity was present.
The remainder was dialyzed for 20 hours at 2' versus 1 liter of 1 mM NaCl, with replacement of outside medium at hours 2 and 16, and the dialyzed, carboxymethylated protein was assayed for radioactivity in a scintillation counter and analyzed for DTNB reactivity (see No. 4 of this  table).
All protein-bound radioactivity was assumed to be in the form of S-carboxymethylcysteine (Cm-cysteine). Similar results were obtained when the incubation with iodoacetic acid contained NAD+ (0.12 pmole) or 2-mercaptoethanol (4 pmoles) or both. b Results were similar with native enzyme or NAD+-loaded enzyme (see "Experimental Procedure"). c Buffer A, Tris-EDTA buffer; Buffer B, Buffer A plus 5 M guanidine hydrochloride or 8 M urea; Buffer C, Buffer B plus 2mercaptoethanol (amount equimolar with iodoacetic acid). d Enzyme (0.04 pmole) was incubated for 30 min at 2' with iodoacetic acid (2.3 pmoles) in 1 ml of Buffer A. Solid guanidine hydrochloride or urea was then added, with (No. 3) or without (No. 2) 2.3 pmoles of concentrated 2-mercaptoethanol, and incubation was continued for 4 hours at 23". The mixture was then dialyzed for 16 hours at 2' versus 1 liter of 1 mM NaCl, with replacement of outside medium at hours 2 and 12; precipitated protein inside the dialysis sac was redissolved by additional dialysis for 4 hours at 2" versus 1 liter of 1 mN HCl.
The protein solution was then taken to dryness in hydrolysis tubes and analyzed for amino acid content as described in Table I. 0 The starting material was dialyzed, carboxymethylated protein from No. 1 of this table (see footnote a). Table III).
We do not understand this differential reactivity of the unfolded polypeptide chains. If a reducing environment was present during the denaturation, the third cysteine became reactive with iodoacetic acid (No. 3 of Table III).
We were unable to detect sulfhydryl reactivity corresponding to the fourth cysteic acid residue of the performic acid-oxidized subunit.

Catalytic Properties
Sulfhydryl and pH Effects-The routine assay procedure for E. coli glyceraldehyde phosphate dehydrogenase, described in the preceding paper (I), included a previous incubation step with 3 InM 2-mercaptoethanol.
Without this precaution specific activity of the purified enzyme was low. Fig. 5 illustrates the sulfhydryl activation effect as a function of 2-mercaptoethanol concentration in the assay mixture. A similar sulfhydryl requirement has been observed with yeast glyceraldehyde phosphate dehydrogenase (38). The standard assay mixture contains 50 InM sodium pyrophosphate buffer at pH 8.5. Use of other buffers yielded the following pH effects.
With 50 mM sodium pyrophosphate buffers activity was maximal at pH 8.8 (40 i. u. per mg of protein) with 85% maximal activity at pH 8.3 and at pH 9.3. With 50 InM Tris-chloride buffers activity was maximal at pH 8 (42 i. u. per mg of protein) with 85% maximal activity at pH 7.5 and at pH 8.5. With 50 mu triethanolamine-chloride buffers activity was maximal at pH 8 (53 i. u. per mg of protein) with 75% maximal activity at pH 7.5 and at pH 8.5.
Effects of Nucleotides on Enzymatic Activity-Two types of effects on the dehydrogenase activity of yeast glyceraldehyde phosphate dehydrogenase have been described when adenine nucleotides were incubated with this enzyme, which, like the E. coli protein, contains little or no bound NAD+ (38). Stance1 and Deal have reported that prolonged incubation at 0" with ATP caused progressive inactivation of the enzyme concomitant with dissociation of the protein (39,40). Yang and Deal have also described an instantaneous inhibitory effect of adenine nucleotides present in enzyme assay mixtures (41). When E. coli glyceraldehyde phosphate dehydrogenase was mixed with a 200to 500-fold molar excess of either ATP or ADP, there was 50yo inactivation by 6 hours,i an effect similar to that reported with the yeast enzyme (39). However, inclusion of ATP or ADP in the E. coli enzyme assay mixture at concentrations up to five times that of NAD+ did not alter activity in the slightest.
Thus the E. coli protein seems to show the long term but not the short term inhibitory effects of ADP and ATP. Park and her associates have documented an esterase activity of yeast glyceraldehyde phosphate dehydrogenase as manifested by catalytic hydrolysis of p-nitrophenyl acetate (26). The same active site cysteine (corresponding to cysteine-149 in the mammalian enzymes) that is carboxymethylated in the iodoacetic acid reaction undergoes acetylation during the esterase reaction (42). As expected, iodoacetate abolishes this activity. Table  IV summarizes results with the E. coli enzyme.
Whether or not the enzyme was first treated with charcoal, there was catalytic hydrolysis of p-nitrophenyl acetate at a level comparable to those reported for the yeast and charcoal-treated rabbit muscle enzymes (26). This activity was inhibited by previous treatment of the protein with iodoacetic acid (footnote a to Table IV). As 1 Enzyme (2 to 5 PM) was incubated with ATP or ADP (1 mrd) in 20 rnM N-tris(hydroxymethyl)methyl glycine buffer, pH 7.8, at 0" for 6 hours.
Aliquots were then assayed for dehydrogenase activity.
Control mixtures lacking nucleotides showed no decay of activity.  (3).
Two general groups of glyceraldehyde phosphate dehydrogenases can be distinguished on the basis of NAD+ content in the isolated protein, those which contain substantial amounts of bound NAD+ and those with little or no attached coenzyme. The E. coli enzyme falls in the latter group, which also includes the enzymes from yeast (38), sturgeon, turkey, and pheasant (2).
However, the E. coli protein apparently will not crystallize unless it is first loaded with NAD+ ( Fig. 4), whereas the other four proteins readily form crystals essentially devoid of cofactor (2,38). In this respect E. coli glyceraldehyde phosphate dehydrogenase resembles the rabbit muscle enzyme from which NAD+ has been removed by charcoal treatment (43). Although we have not attempted to measure NADf-binding constants with the E. coli protein, and it is likely that there are cooperativity effects as with other glyceraldehyde phosphate dehydrogenases (5, 36, 44)) the described ion exchange, gel filtration, and dialysis experiments with artificially NAD+-loaded protein indicate that the purified E. coli enzyme has retained the capacity to strongly bind coenzyme. While it is possible that one or more of the procedures used in purification of the enzyme (1) may have removed previously bound coenzyme, this seems unlikely, and the explanation for the relatively NADf-free state of the isolated protein probably must be sought elsewhere.
The problem may be explicable simply in terms of thermodynamic association between NADf and apoprotein at their respective levels in the E. coli cell compartment where they coexist (although such cell compartments would likely be disrupted during preparation of cell extracts (l), allowing exposure of apoprotein to the substantial levels of NAD+ which are present in E. coli (45)). Alternatively, there may be activities or factors in E. coli which regulate the association or rate of association between NAD+ and glyceraldehyde phosphate dehydrogenase apoprotein. 'I'he differential sulfhydryl reactivity of the cysteine residues in this protein is confusing (Table III) and diflicult to relate to the sulfhydryl sensitivity of enzymatic activity (Fig. 5). Inability to detect sulfhydryl reactivity corresponding to the fourth cysteic acid residue of the performic acid-oxidized enzyme subunit (Table I) is especially puzzling.
There is precedent for chemical masking of sulfhydryl residues in bacterial enzymes (e.g. streptococcal proteinase (46)), and it is possible that a similar phenomenon may account for the discrepancy observed in the present case. It is also possible that the cysteic acid analyses of performic acid-oxidized protein were in error, although we have performed these measures in duplicate for each of two hydrolysis times ( Table I).
Resolution of this problem will probably have to await detailed analyses of the peptide sequence of the enzyme subunit.
We hope that these preliminary studies of a bacterial glyceral-dehyde phosphate dehydrogenase will complement the growing body of information which is being accumulated on this important glycolytic enzyme.
Structural studies are now well advanced with the proteins from pig and lobster muscle and from yeast (4) .2 Documentation of the basic tetrameric structure and